Partial Discharge Monitoring and Diagnosis of Power Generator

Transcription

1 Partial Discharge Monitoring and Diagnosis of Power Generator Gao Wensheng Institute of High Voltage & insulation tech. Electrical Eng. Dept., Tsinghua University

2 Currently preventive maintenance is widely used. But it is carried out off-line. The shortcoming is as follows: Be carried out at de-energized status. However, it is difficult for most power equipment to stop the operation in normal condition. Be carried out periodically. But it is still possible for power equipment to initiate fault during the testing intervals. -4-

3 Sometimes it is not necessary to make the test and maintenance of equipment between regular intervals. The waste is tremendous. The equipment condition when power is off does not reflect the insulating condition truly, thus the identifying accuracy will be influenced. In the traditional PD test, the test voltage possibly can create new fault in power generator. -5-

4 Condition Based Maintenance Insulating performance Condition monitoring Fault diagnosis Maintenance suggestion -6-

5 Partial Discharge Partial Discharge is a discharge in a gas-filled void or on a dielectric surface. Partial discharges are the major cause of progressive failures in HV insulation and have drastic effects on the ageing process of HV generator resulting in early breakdown. -7-

6 Partial discharge (PD) important phenomenon Generators / transformers large capacity Affect of faults much more serious Great attention operational reliability Insulation faults possess a great part -8-

7 Partial Discharge Monitoring A discharge monitoring system for successful early detection of power generator breakdown. It is an essential task in modern asset management of power system. -9-

8 Equipment On-line monitoring system of generator Standard pulse Generator A C 0 A Transformer Neutral B B C C CT 4 C 3 C 2 C 1 C 1 ~ C 3 : Capacitance CT 3 CT 2 CT CT 1 1 ~ CT 4 : Sensors -10-

9 Main Exciter Grnerator Subexciter Outlet Bushing IPB Factory Building Creeping Flange Creeping Flange Factory Building A - IPB C - IPB B - IPB CT Transformer CT Transformer CT Transformer Wall CT CT CT CT Wall CT CT -11-

10 Foundations of PD monitoring technologies 1. Noise rejection 2. PD type recognition 3. On-line diagnosis -12-

11 1. Noise rejection Partial Discharge in power equipment Electric charge moving Electromagnetic radiation Ultrasonic amply vibration PD current transducer UHF PD Sensor Ultrasonic PD Sensor PD monitoring system -13-

12 The difficulty of condition monitoring and fault diagnosis is to depress the strong electromagnetic interference on-site. The different hardware and software filters are used : High-pass filter based on Remez algorithm could be used to real-time suppress the periodical narrow-band noise. Wavelet de-nosing method could reduce background noises effectively. -14-

13 Noise rejection result (a) Original detection signal (b) Signal after de-noise -15-

14 Noise Rejection Method based on UHF 220 V T 1 T 2 T 3 R C x Antenna C C K z 1 C z 2 Tektronix TDS-684C C T -16-

15 In comparison with HF detection, the UHF detection method had favorable anti-interference characteristic but its disadvantage was unable to get the actual discharge quantity by calibration. Therefore a new method for the suppression of pulse-shaped noise was developed, this combined de-noising method based on the fusion of UHF and HF measurements so that fully utilized the favorable anti-interference in UHF measurement and calibrating ability in HF measurement. -17-

16 a. time window b. after extraction in time-domain -18-

17 2. PD type recognition Basis judgments criterion of fault is apparent quantity of PD and its developing trend. Furthermore, we need to know the fault types of PD sources for maintenance suggestions. -19-

18 Typically Partial Discharge insulation HV shield HV lead board HV float Ground Surface discharge Lead discharge Ground Ground Floating discharge conductor High resistance anti-corona layer low resistance anti-corona layer main insulation layer model bar -20-

19 PD current detection result -21-

20 The partial discharge waveform will be distorted after the propagation along the winding, Thus it is difficult to estimate the discharge intensity and to recognize the fault types. -22-

21 We obtained discharge fingerprints, which were represented by T-F spectrum and ϕ-q-n graphs. A fingerprint bank of typical discharges in power equipment was established. Wigner Ville T-F Spectrum ϕ -Q-N 3D distribution charts -23-

22 With the 3D distribution charts the fault patterns and severity could be recognized based on ANN. x 0 f(x) A f(x) B x 1 f(x) A f(x) B f(x) C y x 2 f(x) A f(x) S 1 S 2-24-

23 3. On-line diagnostic expert system for generators For the analysis of fault modes about generators, the fault tree analysis (FTA) method was developed. Diagnostic fault trees of each subsystem were obtained which could supply with the theoretical base for the production diagnostic rule-base. The automatic diagnostic mode and man-machine interactive diagnostic mode are used to evaluate the state of generators based on the knowledge-based system (KBS). -25-

24 Architecture of knowledge-based system ( KBS ) Visual Record Heuristic Knowledge Refining Knowledge Preventive Test Data KB PD Test Data Generator Document Routine Test Data Standards -26-

25 Expression of Knowledge Rule name: Stator Over-current Premise: (Current of Phase A Exceeding 1.2 times Rating Current with CF 1 =0.6) OR (Current of Phase B Exceeding 1.2 times Rating Current with CF 2 =0.6) OR (Current of Phase C Exceeding 1.2 times Rating Current with CF 3 =0.6) Conclusion: 1. Generator Over-load with CF 4 = Short Circuit or Vibration CF 5 =

26 Synthesis database system ( SDBS ) Historical Database Diagnostic Fact-base SDBS Dynamic Database Fault Case-base -28-

27 Fault diagnostic expert system ( FDES ) SCADA Runtime Data DAU Case-base Diagnostic Facts Historical Database Knowledge Base Inference Machine Management Module Result Diagnosis Dynamic Database Man-Machine Graphic User Intreface -29-

28 Virtual Electrical Hospital -30-

29 Further Work Developing an artificial intelligent system for comprehensive assessments to power generator. Based on the data collected from monitoring, power generators residual life could be assessed. Collect data from local monitoring units (RTU) and send the data to central control room using wireless technology. -31-

30 Supported by National Natural Science Foundation of Beijing (Project No ) and National Natural Science Foundation of China (Project No ) -32-